The major part of our code is written in Julia 1.7.0. The RL part is implemented with Python 3.9 and PyTorch.
After installing Julia 1.7, enter its REPL from this root directory and type the following two commands to set up the proper Julia environment.
]activate .(activate the current environment)instantiate(install all dependencies of this project)
- The implementation of the core functions including MILP and IL is provided in the HEMS package in ./src/HEMS. All functions are documented in details.
- Jupyter notebooks in ./notebook are executed to run different experiments with the above HEMS package. The notebooks can be run in VS code or in a browser with IJulia. Alternatively, you may view the notebooks directly on GitHub.
- Run notebook\IL\IL_data.ipynb to generate the expert data.
- Run notebook\IL\BC_shiftable.ipynb to train the DNN agents for shiftable loads via IL based on the above expert data.
- Run notebook\IL\inspection.ipynb to inspect characteristics of the algorithms as shown in the paper.
- The home configuration and scenario data are stored in the ./data directory.
- The other directories contain intermediate images, data, and models etc.
- Code for RL algorithms is available in a separate repository: https://github.com/handworker222/HEMS_RL.